# Project Report: Building a Smart Pong Opponent - From Simple Rules to Adaptive Behavior 
## Motivation Pong is one of the **simplest yet most iconic video games** ever created. Two paddles, a ball, and a goal: simple rules, but surprisingly engaging gameplay. Beyond nostalgia, Pong is also a great playground for exploring basic AI and adaptive behavior. In this study, we explore how a simple computer opponent can “learn” to respond to the player by adjusting its paddle position relative to the ball. This kind of **simple rule-based intelligence** illustrates the foundation of more complex agent behavior in games, simulations, and robotics. ## The Model: Adaptive Pong Opponent The game is classic Pong: a ball bounces off walls, the player controls one paddle, and the computer controls the other. The twist comes in the AI of the computer paddle. **Instead of moving randomly or following a fixed pattern**, the computer updates its y-position relative to the ball’s y-position. In pseudo-logic: - Determine the vertical distance between the ball and the computer paddle. - Move the paddle up or down proportionally to reduce the distance. - Optionally, include a maximum speed so the paddle cannot move instantly to the ball’s position. Despite the simplicity, this creates an opponent that **appears intelligent**: it “tracks” the ball, reacts to the player’s moves, and can even “miss” if the speed is limited appropriately. ## Play the Game Try out the adaptive computer opponent here: ▶ Play Pong vs. Computer: [Play](/assets/pong/games/ai/index.html) The computer paddle moves automatically, and you control your paddle with Up/Down arrow keys. Against Another Human You can also play Pong against another human on the same keyboard: ▶ Play Pong vs. Human: [Play](/assets/pong/games/pvp/index.html) Control the left paddle with W/S and the right paddle with Up/Down arrow keys. This setup allows two players to compete in real time on a single keyboard. ## Implementation Note The core logic of the adaptive AI is simple, based on tracking the ball’s vertical position. With a small speed limit, the computer opponent becomes challenging but beatable, making it an excellent starting point for experimenting with agent-based behavior and simple adaptive rules.